Transcriber

What it does: converts speech in an audio or video file into text, with timestamps. It can export subtitles and transcripts in several formats, and can even translate speech into English. All transcription happens on your device.

How to use it

1. Open Transcriber.
2. Drop audio or video files onto the panel, or click to browse. Common audio and video types are supported (for example MP3, WAV, M4A, MP4, MKV, MOV, and more). You can add several files to transcribe as a batch.
3. Choose a Model (see below).
4. Choose your output formats — you can select more than one:
   • SRT and VTT — subtitle files with timestamps.
   • TXT — plain text transcript.
   • JSON — structured data with timing details.
5. Set the options you want (below), then click Transcribe. Each file becomes a job.

Options explained

• Model: transcription models trade off speed against accuracy. Smaller models are faster and use less disk space; larger models are more accurate but slower. Each model is downloaded on demand the first time you use it and verified before use, then kept for next time. The picker shows each model's size and a speed/accuracy guide, and marks the ones you've already downloaded.
• Word-level timestamps: adds a marker for each individual word (more precise, but slower). Off means timestamps are per phrase/segment.
• Language: choose the spoken language, or leave it on automatic detection. Some smaller, English-only models always transcribe as English.
• Translate to English: transcribe non-English speech directly into English text.

Output

Saved to the Transcripts subfolder, one file per selected format.

Tips & notes

• The first time you use a given model, expect a short one-time download before transcription begins. You can start the download yourself from the model picker.
• Larger models noticeably improve accuracy on difficult audio (accents, background noise); try a bigger model if a result isn't accurate enough.